A Method of Video Flame Detection Based on Multi-Feature Fusion

A method of video flame detection based on multi-feature fusion is presented in this paper. Physical characteristics of flame, including color clues, flame movement and flame flicker are incorporated into the scheme to detect fires in color video frames. Firstly, mean filtering was used to smooth the video frames and a flame color filtering algorithm was adopted to extract flame candidate regions from video frames. Secondly, detected flame candidate regions were categorized into dynamic candidate regions and static non-candidate regions by using XOR algorithm. Finally, a flame flicker identification algorithm based on flame brightness variation was used to extract true flames from dynamic flame candidate regions. Experiments show that the proposed method is effective and robust which remains with strong anti-disturbance ability.